mAPKL (version 1.4.2)

classification: Classify samples according to the SVM algorithm

Description

This function performs classification through the Support Vector Machines (SVM) algorithm. The algorithm applies on the "exemplars" dataset. It produces a classification result either on the training set or on a validation set. This function estimates how well the selected "genes" from mAP-KL method discriminate between two phenotypes.The default SVM settings are: "linear" kernel and 5-folds cross-validation. Regarding the parameters for the "linear" kernel ,cost parameter, and for the "radial" kernel, cost and gamma parameters, are estimated automatically through the tune.svm function as described in e1071 r-package.

Usage

classification(trExemplObj,classLabels,valExemplObj=NULL,kf=5,kernel="linear")

Arguments

trExemplObj

The exemplars train eSet object.

classLabels

The varLabels name in the eSet object where the class labels are stored e.g "type".

valExemplObj

The exemplars validation eSet object (if not NULL).

kf

The k-folds value of the cross-validation parameter.The default value is 5-folds. By setting "Loo" or "LOO" a Leave-One-Out Cross Validation is performed

kernel

The type of kernel used for the classification analysis.The default kernel is "linear"

Value

• classL
  The labels of the train set
• valClassL
  The labels of the validation set if not NULL
• predLbls
  The predicted labels according to the classification analysis

Examples

library(mAPKLData)
data(mAPKLData)
breast <- sampling(Data=mAPKLData, valPercent=40, classLabels="type", seed=135)
normTrainData <- preprocess(breast$trainData)
normTestData <- preprocess(breast$testData)

exprs(breast$trainData)<-normTrainData$clL2.normdata
exprs(breast$testData)<-normTestData$clL2.normdata

out.clL2 <- mAPKL(trObj=breast$trainData, classLabels="type",
valObj=breast$testData,dataType=7)

clasPred <- classification(trExemplObj=out.clL2@exemplTrain, classLabels="type",
valExemplObj=out.clL2@exemplTest)